1 / 9

Performance: Parallel Each

Performance: Parallel Each. David Liebtag. Title slide. What is performance?. How many calculations are performed? How much storage is used? How much data is read and written? How much data is processed? How fast an application runs? How accurate are the application’s answers?.

carol
Download Presentation

Performance: Parallel Each

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Performance: Parallel Each David Liebtag Title slide

  2. What is performance? How many calculations are performed? How much storage is used? How much data is read and written? How much data is processed? How fast an application runs? How accurate are the application’s answers? Basic text slide

  3. How does hardware affect performance? Operation and clock speed Cache size Memory size Virtual paging Number of processors Number of machines Basic text slide

  4. What affects distributed computing performance? Detecting opportunities to distribute processing Slicing the data Copying the data to the processors Number of processors Copying the results Minimizing the number of times operations are distributed Basic text slide

  5. What affects interpreter performance? Scalar operations →BIN/LABEL Array primitives Z←A×B Multiple primitive loops Z←A×B×C Array operators Z←FOO¨A Basic text slide

  6. Parallel Each PEACHP - Parallel Each using processors PEACHT - Parallel Each using threads Parallel each operators start multiple APL2 interpreters Functions applied in separate workspaces with no side-effects Basic text slide

  7. Parallel Each Syntax result ← [larg] (function PEACHP (options [processors])) rarg result ← [larg] (function PEACHT options) rarg function Character vector, name of external function options Vector of character vectors, APL2 invocation options processors Integer vector, processor identifiers, default based on CPUCOUNT larg and rarg Each arguments Basic text slide

  8. Let’s try some things,,, Basic text slide

  9. Room for Improvement Avoid copying data Use multiple threads for primitive array operations Coalesce loops – idiom recognition Rank operator Basic text slide

More Related